Cloud Cruiser became HPE Consumption Analytics on Nov. 1st. You'll still see the old name in places while we update this site.

 

 

Consumption Analytics Documentation

Home > Cloud Cruiser 4 > Collecting, transforming, and publishing > VMware vSphere > Performance Collector

Performance Collector

The Performance Collector collects historical performance data for entities collected by the vSphere Snapshot Collector. The performance data for an entity is called a counter. You choose which counters to collect for each entity type that you're collecting with the Snapshot Collector.

This section contains the following articles:

For the recommended steps for using the Performance​ Collector along with the Snapshot​ Collector, see VMware vSphere.

Sampling period

The Performance Collector writes one CC Record for each sample it retrieves from vSphere. Because vSphere samples at several different intervals, the granularity of data available to you in HPE Consumption Analytics Portal is determined by which of these samples you collect. You can select one of these Sampling Periods in the collector properties. See Snapshot Collector for more information.

  • 300 (5 minutes)
  • 1800 (30 minutes)
  • 7200 (2 hours)
  • 86400 (24 hours)

This setting works together with the Aggregation Interval to determine the timeframe of each CC Record written by the collector. For example, if you want records that each span an hour, you should set the Sampling Period to 1800 and the Aggregation Interval to HOURLY. This retrieves 30-minute samples and then aggregates them to create hourly output records.

For more information about historical intervals, see the VMware vSphere API Reference.

For information about configuring the collection interval in vSphere, see the VMware vSphere 4 and vSphere 5 documentation.

Aggregating performance data

In version 4.1 and later, you can make the collector aggregate data by hourly or daily periods. This is different from an Aggregate Rows step in a data worksheet—this aggregation is performed by the collector before it writes CC Records. The most common use for this aggregation is to create hourly records for performance data.

You aggregate by providing an Aggregation Interval for the collector and a Function for each counter you collect. For each counter, the collector performs the selected Function (such as AVG or SUM) on the samples for that counter to aggregate them into a single record for the selected Aggregation Interval.

To aggregate, you must set Aggregation Interval greater than or equal to Sampling Period so that samples are small enough to aggregate. For example, a collection with an Aggregation Interval of HOURLY and a Sampling Period of 7200 (2 hours) will not run.

For information about Function, see Advanced configuration.

vSphere Statistics Level

Each vSphere performance counter specifies a required statistics level, also known as a collection level, for both aggregated data and per-device data. In counter reference files generated by the HPE Consumption Analytics Portal vSphere Performance Collector, these required levels are in the level and perDeviceLevel values in columns M and N. vSphere collects performance data only for the counters whose required aggregated statistics level is met by vSphere's Statistics Level.

If vSphere's ​current Statistics Level isn't high enough to meet the levels required by the performance counters you want to collect, you can address this either by raising the Statistics Level for the interval you're collecting or by lowering the required levels for your desired counters. The following sections explain each of these methods.

Modifying the current vSphere level

This is the easier method, though it can affect the performance of your vCenter server.

To modify the vSphere Statistics Level

  1. Log in to your vCenter server using the vSphere client.
  2. Choose Administration > vCenter Server Settings > Statistics.
  3. In the table, select the interval that equals the samplingPeriod specified in your feed configuration, then click Edit.
    For example, if your samplingPeriod is 7200 seconds, select 2 Hours.
  4. Select a Statistics Level, then click OK twice to save your changes and return to the vSphere window.

Modifying the required level for a counter

It is possible to set the required statistics level for one or more individual performance counters, allowing you to collect performance data without raising the vSphere Statistics Interval. For information about this technique, see this third-party blog post and this VMware knowledge base article.

Collector properties

This section explains the Performance Properties that you set in the Edit Collection wizard.

Before you create a Performance collection, create a corresponding Snapshot collection.

If you don't, when you try to finish the collection wizard or open its Advanced Configuration, you'll see the error message Associated content not found for the collection. This means that the collector could not find the entities file it needs to retrieve sample data. You won't see sample data for the vSphere Performance collection, and will continue to see this error message, until you create a vSphere Snapshot collection in the same workbook for the same data source.

​You can set the following properties:

  • Sampling Period: The interval, in seconds, for which you are collecting performance data. For more information, see Sampling period.
  • Aggregation Interval: (version 4.1 and later) The interval by which to aggregate retrieved performance data before writing CC Records. For more information, see Aggregating performance data.

Advanced configuration

This setion explains how to use the Advanced Configuration dialog box for a VMware vSphere Performance collection. Performance profiles were added in version 4.1. For earlier versions, there is no need to use the Advanced Configuration dialog box.

Input Settings

The properties on the Input Settings tab are also available in the basic configuration. For information, see Collector properties.

Performance Profiles

For each type of managed object that you want to collect performance data for, you must define a performance profile to specify the counters to collect.

You can only collect performance data for entities for which you have already collected snapshot data, so a performance profile for an object type that doesn't also have a snapshot profile has no effect. For more information, see Tracking entities for performance collection.

To create a performance profile

  1. In the Performance Profiles tab, click a Plus icon icon to create an empty profile, then enter the managed object name in the Entity Type field.
    For example, VirtualMachine. The order of the profiles does not matter.
  2. With that profile selected, click Add Row in the pane below, then complete the four fields in this new row in the Counters table.
    Together, the Group, Name, and Rollup Type fields uniquely identify a counter. To determine the counters available for the managed objects you are collecting, use the list of performance counters in the vSphere API Reference. A counter entry in HPE Consumption Analytics Portal consists of the following fields:
    • Group: The counter group to which the desired counter belongs. For example, the cpu group includes CPU utilization counters for VMs, hosts, and other entities.
    • Name: The name of the desired counter. For example, the cpu group includes the totalCapacity and idle counters.
    • Function: (optional) The aggregation function to apply to the samples returned, if the collection's Aggregation Interval is not NONE. You can select MINMAXAVG, SUM, and NONE. For more information, see Aggregating performance data. HPE recommends that you align this function with Rollup Type to produce consistent aggregation from narrower to broader time intervals.
    • Rollup Type: The type of vSphere data rollup to request for the counter within the collection's Sampling Period. You can select AVERAGEMAXIMUMMINIMUMSUMMATION, ​LATEST, and NONE.​ The vSphere documentation shows which rollup types are available for each counter.
  3. Repeat step 2 for each counter you want to collect for this entity type.

Each counter you specify in a performance profile generates an output CC Record for each entity of the given type. The counter is the one measure in the record, named <group>.<name>. For example, cpu.totalCapacity. The record contains four dimensions: EntityType, EntityID, EntityName, and Instance.

The cpu.usage counter is intended to represent the percentage of CPU being used by the VM, but as of vSphere 5.1 the counter value provided is that percentage multiplied by 100. For example, a cpu.usage value of 1161 means CPU usage of 11.61%. HPE recommends that you divide the collected value by 100 or 10,000 in your ETL flow, depending upon how you want to represent it in reports.

Last modified

Tags

Classifications

This page has no classifications.
© Copyright 2018 Hewlett Packard Enterprise Development LP